Vincent Purcell

Data Mining 2 - Final Project

Set up MongoDB Connections

import pymongo

# MongoDB Connections
client = pymongo.MongoClient('mongodb://localhost:27017')
db = client['MLB_DB']

def getStatAvg(stats_list, projections_str, player_projected_stats_doc):
    stat_averages = [0] * len(stats_list)
    count = 0
    for player in player_projected_stats_doc[projections_str].values():
        stat_index = 0
        for stat in stats_list:
            try:
                stat_val = float(player[stat])
            except:
                stat_val = 0
            stat_averages[stat_index] = stat_averages[stat_index] + stat_val
            stat_index = stat_index + 1
        count = count + 1
    stat_averages = [i/count for i in stat_averages]
    return stat_averages

import datahelp

def getData(summaryDB, PlayerDataDB, PlayerProjectionsDB):
    columns = datahelp.columns
    data = []
    for document in summaryDB.find():
        game_id = document['game_id_num']
        for player_data in PlayerDataDB.find({'game_id_num': str(game_id)}):
            home_umpire = [player_data['umpires']['home']['id']]
        for player_projections in PlayerProjectionsDB.find({'game_id_num': str(game_id)}):
            # Get Stat averages
            home_batting_avg = getStatAvg(datahelp.batting_stats, 'home_batting_projections', player_projections)
            away_batting_avg = getStatAvg(datahelp.batting_stats, 'away_batting_projections', player_projections)
            home_pitching = getStatAvg(datahelp.pitching_stats, 'home_pitcher_projections', player_projections)
            away_pitching = getStatAvg(datahelp.pitching_stats, 'away_pitcher_projections', player_projections)
        home_win = [1 if document['home_score']>document['away_score'] else 0]
        data_row = home_batting_avg + away_batting_avg + home_pitching + away_pitching + home_umpire + home_win
        data.append(data_row)
    return data

gameSummaryDB = db['GameSummary']
gamePlayerDataDB = db['GamePlayerData']
playerProjectionsDB = db['PlayerProjectedStats']

data = getData(gameSummaryDB, gamePlayerDataDB, playerProjectionsDB)

import numpy as np
import pandas as pd
import datahelp

columns = datahelp.columns

data_df = pd.DataFrame(np.array(data), columns=columns)
print("Data count (rows, columns) : " + str(data_df.shape))
data_df.head()

Data count (rows, columns) : (1202, 98)

	home hr avg	home gidp avg	home sac avg	home rbi avg	home tb avg	home slg avg	home avg avg	home bb avg	home ops avg	home hbp avg	...	away pitching ip	away pitching w	away pitching r	away pitching pa	away pitching cg	away pitching gs	away pitching ibb	away pitching er	umpire_id	home win
0	15.222222	9.444444	4.000000	80.666667	283.111111	0.393667	0.259111	60.111111	0.715444	10.888889	...	142.1	10.0	65.0	617.0	0.0	11.0	1.0	58.0	484499.0	0.0
1	16.777778	8.333333	3.888889	77.444444	273.888889	0.384889	0.250778	59.555556	0.701222	14.666667	...	57.2	1.0	36.0	265.0	0.0	10.0	1.0	35.0	482666.0	0.0
2	16.333333	7.777778	2.777778	63.222222	214.555556	0.377667	0.230000	45.111111	0.673111	10.888889	...	191.2	7.0	138.0	869.0	0.0	28.0	1.0	130.0	427315.0	0.0
3	18.555556	8.888889	3.222222	78.222222	265.444444	0.406556	0.254111	52.666667	0.722778	12.666667	...	388.2	22.0	185.0	1654.0	0.0	64.0	2.0	168.0	428442.0	1.0
4	17.555556	9.888889	3.222222	81.444444	280.888889	0.409444	0.262444	55.222222	0.728667	9.222222	...	255.1	12.0	149.0	1143.0	0.0	46.0	2.0	138.0	427044.0	1.0

5 rows × 98 columns

from sklearn.model_selection import train_test_split
import seaborn as sns
import matplotlib.pyplot as plt

X = data_df.drop(['home win'], axis = 1)
y = data_df['home win']

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size = 0.15, random_state = 15)
# visualize number of digits classes
plt.figure(figsize=(5,5))
g = sns.countplot(y_train, palette="icefire")
plt.title("Number of Wins and Losses")
y_train.value_counts()

X_train.head()

C:\Users\vin99\AppData\Local\Programs\Python\Python38\lib\site-packages\seaborn\_decorators.py:36: FutureWarning: Pass the following variable as a keyword arg: x. From version 0.12, the only valid positional argument will be `data`, and passing other arguments without an explicit keyword will result in an error or misinterpretation.
  warnings.warn(

	home hr avg	home gidp avg	home sac avg	home rbi avg	home tb avg	home slg avg	home avg avg	home bb avg	home ops avg	home hbp avg	...	away pitching h	away pitching ip	away pitching w	away pitching r	away pitching pa	away pitching cg	away pitching gs	away pitching ibb	away pitching er	umpire_id
976	22.222222	8.777778	2.444444	83.444444	273.555556	0.388667	0.241889	60.000000	0.693444	5.888889	...	85.0	91.2	5.0	43.0	377.0	0.0	13.0	0.0	42.0	521251.0
136	37.666667	8.666667	2.555556	113.777778	357.444444	0.483111	0.264889	78.000000	0.813000	6.777778	...	34.0	35.2	3.0	23.0	162.0	0.0	7.0	0.0	21.0	503493.0
1092	11.000000	5.666667	0.444444	35.666667	117.444444	0.371222	0.220444	29.111111	0.663444	4.333333	...	43.0	51.2	3.0	25.0	215.0	0.0	10.0	0.0	20.0	427139.0
981	28.000000	6.222222	2.000000	100.222222	313.000000	0.442778	0.253667	68.777778	0.770778	8.666667	...	252.0	233.0	11.0	155.0	1024.0	0.0	41.0	0.0	148.0	547380.0
620	33.777778	13.777778	3.666667	129.000000	393.555556	0.434333	0.262778	93.555556	0.770889	10.555556	...	214.0	209.2	10.0	116.0	912.0	0.0	38.0	1.0	109.0	427128.0

5 rows × 97 columns

from sklearn.neural_network import MLPClassifier
from sklearn.metrics import classification_report, confusion_matrix

mlp = MLPClassifier(hidden_layer_sizes=(50,75,98), max_iter=2000)
mlp.fit(X_train,y_train)
y_predict = mlp.predict(X_test)

cm = np.array(confusion_matrix(y_test, y_predict, labels=[1,0]))
confusion = pd.DataFrame(cm, index=['home_win', 'home_loss'],
                         columns=['predicted_win','predicted_loss'])
sns.heatmap(confusion, annot=True)

<AxesSubplot:>

from sklearn.svm import SVC
svc_model = SVC()
svc_model.fit(X_train, y_train)
y_predict = svc_model.predict(X_test)

cm = np.array(confusion_matrix(y_test, y_predict, labels=[1,0]))
confusion = pd.DataFrame(cm, index=['home_win', 'home_loss'],
                         columns=['predicted_win','predicted_loss'])
sns.heatmap(confusion, annot=True)

<AxesSubplot:>

import shap
from sklearn.svm import SVR
svr_poly = SVR(kernel='poly', C=1e3, degree=2)
svr_poly.fit(X_train, y_train)
y_predict = svr_poly.predict(X_test)
explainer = shap.explainers.Permutation(svr_poly.predict, X_test)
shap_values_svc = explainer(X_test[:100], silent=True)
shap.plots.bar(shap_values_svc)

from sklearn.neural_network import MLPRegressor
mlp_regr = MLPRegressor(random_state=1, max_iter=2000).fit(X_train, y_train)
y_predict = mlp_regr.predict(X_test)
explainer = shap.explainers.Permutation(mlp_regr.predict, X_test)
shap_values_svc = explainer(X_test[:100], silent=True)
shap.plots.bar(shap_values_svc)

from sklearn import tree
from dtreeviz.trees import *
import dtreeviz as dtviz


max_depth = 10
clf_dt = tree.DecisionTreeClassifier(max_depth=max_depth)
clf_dt = clf_dt.fit(X_train, y_train)
#plt.figure(figsize=(35,12))
#tree.plot_tree(clf_dt,feature_names=columns,class_names=['loss','win'],rounded=True,fontsize=8);
#plt.show()
predictions_dt = clf_dt.predict(X_test)
accuracy = 1-(((y_test != predictions_dt).sum())/len(X_test))
print("Decision Tree Classification (max depth =" + str(max_depth) + ")")
print("Number of mislabeled points out of a total %d points: %d" % (len(X_test), (y_test != predictions_dt).sum()))
print("Accuracy: %f" % (accuracy))



viz = dtviz.trees.dtreeviz(clf_dt,
                   X_train,
                   y_train,
                   target_name='home_win',
                   feature_names=columns[:-1],
                   orientation="LR",
                   class_names=['loss','win'],
                   fancy=True,
                   X=None,
                   label_fontsize=12,
                   ticks_fontsize=8,
                   fontname="Arial")

viz

Decision Tree Classification (max depth =10)
Number of mislabeled points out of a total 181 points: 59
Accuracy: 0.674033

clf_dt = tree.DecisionTreeClassifier()
clf_dt = clf_dt.fit(X_train, y_train)
#plt.figure(figsize=(100,15))
#tree.plot_tree(clf_dt,feature_names=columns,class_names=['loss','win'],rounded=True,fontsize=7);
#plt.savefig('dtree_high_dpi',dpi=600)
predictions_dt = clf_dt.predict(X_test)
accuracy = 1-(((y_test != predictions_dt).sum())/len(X_test))
print("Decision Tree Classification")
print("Number of mislabeled points out of a total %d points: %d" % (len(X_test), (y_test != predictions_dt).sum()))
print("Accuracy: %f" % (accuracy))

Decision Tree Classification
Number of mislabeled points out of a total 181 points: 69
Accuracy: 0.618785

from sklearn.ensemble import RandomForestClassifier

max_depth = 7
criterion = "entropy"
clf_rf = RandomForestClassifier(max_depth=7, random_state=0, criterion="entropy")
clf_rf.fit(X_train, y_train)
predictions_rf = clf_rf.predict(X_test)

accuracy = 1-(((y_test != predictions_rf).sum())/len(X_test))
print("Random Forest Classification")
print("max depth = " + str(max_depth) + ", criterion = " + criterion)
print("Number of mislabeled points out of a total %d points: %d" % (len(X_test), (y_test != predictions_rf).sum()))
print("Accuracy: %f" % (accuracy))

estimator = clf_rf.estimators_[5]

viz = dtviz.trees.dtreeviz(estimator,
                   X_train,
                   y_train,
                   target_name='home_win',
                   feature_names=columns[:-1],
                   orientation="LR",
                   class_names=['loss','win'],
                   fancy=True,
                   X=None,
                   label_fontsize=12,
                   ticks_fontsize=8,
                   fontname="Arial")

viz

Random Forest Classification
max depth = 7, criterion = entropy
Number of mislabeled points out of a total 181 points: 52
Accuracy: 0.712707

gameSummaryFutureDB = db['GameSummaryFutureGames']
gamePlayerDataFutureDB = db['GamePlayerDataFutureGames']
playerProjectionsFutureDB = db['PlayerProjectedStatsFutureGames']

future_data = getData(gameSummaryFutureDB, gamePlayerDataFutureDB, playerProjectionsFutureDB)

future_data_df = pd.DataFrame(np.array(future_data), columns=columns)
print("Data count (rows, columns) : " + str(future_data_df.shape))

y_test_future = future_data_df['home win']

future_data_df = future_data_df.drop(['home win'], axis = 1)
future_data_df

Data count (rows, columns) : (9, 98)

	home hr avg	home gidp avg	home sac avg	home rbi avg	home tb avg	home slg avg	home avg avg	home bb avg	home ops avg	home hbp avg	...	away pitching h	away pitching ip	away pitching w	away pitching r	away pitching pa	away pitching cg	away pitching gs	away pitching ibb	away pitching er	umpire_id
0	10.222222	2.666667	0.444444	40.555556	111.444444	0.409000	0.236111	28.333333	0.724556	3.777778	...	129.0	103.1	6.0	76.0	474.0	0.0	23.0	0.0	70.0	482641.0
1	11.777778	5.777778	0.111111	33.666667	103.888889	0.379222	0.215889	34.333333	0.683222	2.888889	...	49.0	55.1	5.0	28.0	235.0	0.0	6.0	1.0	24.0	608093.0
2	11.111111	5.333333	0.333333	37.111111	118.222222	0.476444	0.265444	26.555556	0.809000	3.444444	...	126.0	131.0	9.0	64.0	553.0	0.0	23.0	0.0	58.0	427292.0
3	10.333333	6.333333	0.888889	39.444444	125.666667	0.381889	0.247667	26.666667	0.686556	3.000000	...	59.0	68.2	5.0	37.0	285.0	0.0	15.0	0.0	35.0	607884.0
4	10.333333	5.555556	0.444444	33.333333	107.555556	0.403556	0.228778	28.555556	0.709222	2.333333	...	87.0	97.2	3.0	37.0	410.0	0.0	14.0	0.0	34.0	483630.0
5	6.444444	3.000000	0.111111	24.111111	72.888889	0.390222	0.249556	14.666667	0.692222	1.888889	...	23.0	27.1	1.0	14.0	116.0	0.0	5.0	0.0	13.0	521889.0
6	13.555556	6.555556	0.333333	45.333333	130.888889	0.431000	0.244111	29.000000	0.754000	6.777778	...	12.0	18.1	2.0	7.0	77.0	0.0	1.0	2.0	6.0	503490.0
7	11.555556	6.555556	0.666667	36.000000	115.555556	0.453000	0.268889	28.888889	0.800889	2.666667	...	92.0	113.1	6.0	52.0	468.0	0.0	22.0	0.0	51.0	427248.0
8	9.888889	5.444444	1.444444	37.777778	113.222222	0.398667	0.240778	26.111111	0.705000	4.222222	...	73.0	74.1	4.0	39.0	311.0	0.0	15.0	1.0	39.0	503077.0

9 rows × 97 columns

future_predictions = clf_dt.predict(future_data_df)

count = 0
for document in gameSummaryFutureDB.find():
    summary_string = document['away_name'] + ' @ ' + document['home_name'] + "  [" + document['game_datetime'] + "]"
    if future_predictions[count] == 1:
        print(summary_string)
        print("     winner: " + document['home_name'])
        #print("     winner: " + document['home_name'] + "  [proba: " + str(round(future_predictions[count],4)) + "]")
    else:
        print(summary_string)
        print("     winner: " + document['away_name'])
        #print("     winner: " + document['away_name'] + "  [proba: " + str(round(future_predictions[count],4)) + "]")
    count = count + 1

#accuracy = 1-(((y_test_future != future_predictions).sum())/len(future_data_df))
#print("Decision Tree Classification")
#print("Number of mislabeled points out of a total %d points: %d" % ((len(future_data_df)), (y_test_future != future_predictions).sum()))
#print("Accuracy: %f" % (accuracy))

Baltimore Orioles @ Tampa Bay Rays  [2021-08-16T23:10:00Z]
     winner: Tampa Bay Rays
Los Angeles Angels @ New York Yankees  [2021-08-16T23:05:00Z]
     winner: New York Yankees
Oakland Athletics @ Chicago White Sox  [2021-08-17T00:10:00Z]
     winner: Chicago White Sox
Houston Astros @ Kansas City Royals  [2021-08-17T00:10:00Z]
     winner: Kansas City Royals
Cleveland Indians @ Minnesota Twins  [2021-08-17T00:10:00Z]
     winner: Minnesota Twins
Atlanta Braves @ Miami Marlins  [2021-08-16T23:10:00Z]
     winner: Miami Marlins
Chicago Cubs @ Cincinnati Reds  [2021-08-16T23:10:00Z]
     winner: Cincinnati Reds
New York Mets @ San Francisco Giants  [2021-08-17T01:45:00Z]
     winner: San Francisco Giants
San Diego Padres @ Colorado Rockies  [2021-08-17T00:40:00Z]
     winner: San Diego Padres